Subtyping of acute myeloid leukaemia (AML) has made significant progress, exemplified by the recent classification updates by the World Health Organization and International Consensus Classification. AML subclassification is predominantly genetics-based, despite research showing the benefits of transcriptomic profiling on top of known genetic markers. However, a comprehensive survey of subtypes in AML defined by gene expression has yet to be performed. To this end, we integrated mRNAseq data from 1337 patients and five studies, with corresponding biological and clinical data. We defined 19 gene expression-based subtypes, further stratifying AML. We found that KMT2A leukaemias with fusion partners MLLT3, MLLT10 and MLLT1 clustered together, while KMT2A-MLLT4 displayed a distinct gene expression pattern, suggesting differences in their aetiology. We discovered a transcriptional CEBPA subtype, of which only 40% had a CEBPA bZIP indel. Regardless of mutation status, all patients within this CEBPA cluster had the same favourable outcome. We found four NPM1- enriched transcriptomic subtypes, each with distinct co-mutation patterns and associated ex-vivo drug responses. Similarly, we identified nine AML with myelodysplasia-related changes (AML-MRC) subtypes, dividing a subtype making up one-third of the AML patients into novel groups with different outcomes and drug response profiles. In conclusion, we provide an unprecedented overview of the transcriptomic subtypes in AML and illustrate their potential for AML diagnostics.